Gunshot Detection Using Edge ML and Multilateration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current gunshot detection systems face challenges in accurately locating gunshots in urban environments due to high bandwidth requirements, echo issues, and incorrect location identification, especially when recording devices are blocked or sporadically placed.
Innovation Solution
The system employs microphones mounted on light fixtures across a city, using machine learning techniques and multilateration to process audio data and determine the location of gunshots by generating spectrograms and transmitting only relevant data to a remote server, reducing bandwidth usage and improving location accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If recording devices continuously stream audio data to a remote server, then gunshot detection capability is improved, but network bandwidth consumption increases significantly
Solution Approach 1:
The recording devices perform preliminary gunshot detection processing locally before transmitting data. The device analyzes audio data in real-time using machine learning models to identify gunshot sounds, and only transmits detected gunshots or relevant audio segments to the remote server, rather than continuously streaming all audio data.
Solution Approach 2:
The system extracts and transmits only the essential information (detected gunshots and relevant audio segments) from the continuous audio stream, separating the critical detection data from the bulk audio data that remains processed locally.
2Area of stationary object
If multiple recording devices are deployed across the city, then coverage area is improved, but processing power requirements at the remote server increase
Solution Approach 1:
The system segments the processing workload between edge devices and the remote server. Each recording device independently performs local gunshot detection using machine learning models, filtering and preprocessing audio data before transmission. This distributes computational burden across multiple independent nodes rather than concentrating all processing at the central server.
Solution Approach 2:
Each recording device serves itself by performing local audio analysis and gunshot detection using embedded machine learning models. The devices autonomously identify gunshots and prepare data for transmission without requiring centralized processing of all audio streams, enabling independent operation and reducing remote server processing requirements.
3Ease of manufacture
If recording devices are placed sporadically around the city, then deployment complexity is reduced, but location accuracy of gunshot detection deteriorates
Solution Approach 1:
The system uses feedback from multiple recording devices that detect the same gunshot event to improve location accuracy. When multiple devices detect a gunshot, the system analyzes the time differences of arrival and uses this feedback to calculate and refine the gunshot location through triangulation algorithms, compensating for sporadic device placement.
Solution Approach 2:
The system transitions from relying on spatial density of devices to utilizing the temporal dimension for location determination. By analyzing the time of arrival differences of gunshot sounds at multiple sporadically placed devices, the system calculates location in the time-space domain, adding temporal measurement as an additional dimension for location precision.
4Ease of operation
If recording devices are placed at lower heights, then installation ease is improved, but audio recording quality deteriorates due to impedance from passerby objects
Solution Approach 1:
The system changes the installation parameter from ground level to elevated positions on light fixtures. This parameter change raises the recording devices above the impedance zone created by passerby objects, vehicles, and street furniture, improving audio capture quality while maintaining straightforward installation through standardized mounting interfaces on existing light fixtures.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables quick and accurate detection and notification of gunshot locations, allowing authorities to respond promptly and potentially apprehend perpetrators or collect evidence.
Implementation Method 1
A microphone may be mounted within or on the housings mounted to the light fixtures
Implementation Method 2
The processor(s) may execute a machine learning model with the spectrograms to determine whether the spectrograms include audio data of a gunshot
Implementation Method 3
The sound waves of gunshots can be loud and can echo off of buildings surrounding the streets in a city
Data Source
AI summary
An apparatus may include a device housing; a camera; a plurality of microphones inside or mounted to the device housing; and a processor inside the device housing and electrically coupled to the plurality of microphones. The processor can be configured to receive a set of audio data from each of the plurality of microphones; execute a machine learning model using each of the sets of audio data as input to determine whether the set of audio data corresponds to an actionable sound; determine a location of the actionable sound relative to a location of the camera based on a plurality of sets of audio data determined to correspond to the actionable sound; and rotate the camera towards the determined location of the actionable sound.